1.原版
import pickle
import numpy as np
import matplotlib.pyplot as plt
from keras.models import Sequential,Model
from keras.layers import Dropout, Dense, Conv2D, Flatten, MaxPooling2D, Input
from keras.callbacks import ModelCheckpoint
from keras import utils # 归一化变量
from sklearn.model_selection import train_test_split
inpt = Input(shape=(227,227,3))
# (227,227,3) -> (55,55,96) -> (27,27,96)
x = Conv2D(filters=96, kernel_size=(11, 11), strides=(4, 4), activation='relu',kernel_initializer='uniform')(inpt)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (27,27,96) -> (27,27,256)->(13,13,256)
x = Conv2D(filters=256, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (13,13,256)->(13,13,384)->(13,13,384)->(13,13,256)->(6,6,256)
x = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = Conv2D(filters=256, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (6,6,256)->9216->4096->4096->1000
x = Flatten()(x)
x = Dense(4096, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(4096, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(1000, activation='softmax')(x)
model = Model(inputs=inpt, outputs=x)
# 优化器用的sgd
2.改进一下
第一个步长设置为1, padding = same
inpt = Input(shape=(60,60,8))
# (227,227,3) -> (55,55,96) -> (27,27,96)
x = Conv2D(filters=96, kernel_size=(11, 11), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(inpt)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (27,27,96) -> (27,27,256)->(13,13,256)
x = Conv2D(filters=256, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (13,13,256)->(13,13,384)->(13,13,384)->(13,13,256)->(6,6,256)
x = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = Conv2D(filters=256, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu',kernel_initializer='uniform')(x)
x = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(x)
# (6,6,256)->9216->4096->4096->1000
x = Flatten()(x)
x = Dense(4096, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(4096, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(14, activation='softmax')(x)